Optimal Tuning of Kriging Hyper-Parameters Based on Cuckoo Search Algorithm

摘要：

　　Kriging is widely used as a popular type of surrogate models for approximating expensive black box functions in engineering design and optimization.The tuning of kriging hyper-parameters has a direct and influential impact on its approximating capability.It always requires numerical optimization of the nonlinear and multi-modal likelihood function to obtain the optimal kriging hyper-parameters based on the maximum likelihood estimation(MLE)theory.In this work,a cuckoo search based approach is proposed and applied to the optimization of kriging hyper-parameters to ensure and improve the approximation ability of the kriging surrogate model.Firstly,the hyper-parameter optimization model of kriging is built based on MLE.Secondly,CS algorithm is applied on the established optimization model to search the optimum hyper-parameters.Finally,the benchmark test function is adopted as the case to verify the proposed method,and the results of the case study demonstrate that CS outperforms PS,GA and PSO in terms of the quality of the solution and efficiency for optimal kriging hyper-parameter tuning by comparison.

作者: Bo YANG Zhongqi WANG Cheng LI Yonggang KANG Yuan YANG

作者单位: School of Mechanical Engineering,Northwestern Polytechnical University,Xian,710072,China The Ministry of Education Key Laboratory of Contemporary Design and Integrated Manufacturing Technol

会议类型: 国际会议

会议名称: The 2017 2nd International Seminar on Applied Physics, Optoelectronics and Photonics (APOP 2017) (2017年第二届应用物理、光电子学和光子学国际研讨会)

会议地点: 上海

会议语种:英文

页码: 1-8

在线出版日期: 2017-12-30（万方平台首次上网日期，不代表论文的发表时间）